Sign up for Online Meets here. | View Academic Calendar 2023-24 here. | Admissions to BTech 2024 are now open | BTech round 4 application deadline is April 30 | TLF will be relaunched as a Master's program in 2025| Calling talented PhD scholars and postdoctoral fellows in the US to meet us. Register here| Work at Plaksha
Dr. Deepan Muthirayan
Assistant Professor, Plaksha University
Education
PhD, University of California at Berkeley (2016)
Dual Degree (BTech/MTech), IIT Madras (2010)
Bio
Deepan is currently an Assistant Professor at Plaksha University.
He obtained his PhD from the University of California at Berkeley (2016) and Dual Degree (B.Tech/M.tech) from the Indian Institute of Technology Madras (2010). His doctoral thesis work focused on market mechanisms for integrating demand flexibility in energy markets. Before his term at UC Irvine, he was a post-doctoral associate in the department of Electrical and Computer Engineering at Cornell University, where his work focused on optimization, parametric learning and matching markets.
Deepan's current research spans the areas of control systems, topics at the intersection of learning and control, machine learning, online learning and optimization, game theory, mechanism design and their application to cyber-physical systems.
His current research in theoretical machine learning is focused on the following topics:
Online learning: sub-linear regret algorithms for optimization of time-varying dynamical systems, for optimization of non-linear dynamical systems and online learning and optimization in distributed control settings.
Reinforcement learning: (i) Reinforcement learning methods for robust and adaptive control. The importance of this research lies in the fact that such methods are arguably the most critical for addressing the sim-real gap. (ii) A systematic language for reward and training design. The reward and training schemes used currently are specialized even within the space of robotics and within even specific tasks like locomotion. The lack of systematic methods makes the training process very cumbersome. This direction of research aims to address this gap. Application of the above methods to robotics, power systems, and IoT systems.